We’re hoping to improve the inline terminal compatibility soon, but we have noticed the AI can be optimistic about using keyboard interrupts in the terminals!
For now, try adding to your Rules for AI an instruction to not use keyboard interrupts, as this may help avoid this in the short term.
I have done this several times, it’s not helped (though the AI definitely gets more apologetic the more intense I make the rule phrasing when it inevitably interrupts again)
I am also having still having issues with running simple powershell commands. Its very hit or miss.
I’ve mostly been trying to use agent in a helpdesk capacity to connect to azure, add/remove people from groups, this sort of thing. But in the course of testing can reproduce using simple directory listings. One time it’ll work and get the output, the next it’ll say there was some problem with the output and its going to try again even though you can see in the previous commands run it ran correctly.
In some cases you can ask it to look back at the output and it will be able to see that it did output correctly and grab it, othertimes it still won’t see the command output.
same problem here as well, its needing to run several test scripts one after another during testing but sometimes works but sometimes fails. Mostly I’ve seen it fail when its back to back. Sometimes it will try and run a new terminal command and it will run in the previous chat box where the prior terminal command ran - maybe that is helpful to debug.
Same problem.
I instruct it explicitly to not use keyboard interrupts, it apologies a lot, but keeps doing them.
Then, it starts debugging the “expection” (its own keyboard interrupt)
The current status - I cannot get it to run my code, that takes a few minutes to run. Cannot get it to debug itself